Scoring model
Weights are data, not code. Every score stores the model version that produced it, so a decision made eighteen months ago can still be explained.
Factor weights
total 100| Factor | Family | Weight | Availability |
|---|---|---|---|
| Document completenessRequired documents present and readable | Compliance | Live from day one | |
| Certificate validityDays remaining before the earliest expiry | Compliance | Live from day one | |
| Bank verificationAccount confirmed by callback, not by letter alone | Compliance | Live from day one | |
| Category criticalityHow badly a failure here would hurt | Structural | Live from day one | |
| Single-source exposureShare of category spend held by this supplier | Structural | Live from day one | |
| On-time deliveryReceipts on or before the promised date | Performance | Needs history | |
| In-full deliveryQuantity received against quantity ordered | Performance | Needs history | |
| Quality rejectionsUnits rejected at goods receipt | Performance | Needs history | |
| Invoice match qualityClean 3-way matches from ScanMyInvoice | Performance | Needs history |
Weights must total 100 before publishing.
Thresholds and automatic actions
Below this, the supplier is flagged for review at the next sourcing event.
Below this, new purchase orders are stopped until a buyer overrides with a reason.
Versions
- v3 — liveQuality rejections separated from in-full deliveryPublished 02 Jun 2026 by Demo User 1
- v2Single-source exposure added at 10%Published 14 Mar 2026 by Demo User 2
- v1Compliance factors onlyPublished 20 Jan 2026 by Demo User 2
If you publish v4
Nothing is rewritten. Scores keep the version that produced them and are restamped as each supplier is next recalculated.
Suppliers who would rise142
Suppliers who would fall88
Would cross below 456
Six suppliers would fall below the review threshold, two of
them on critical categories. Simulate before publishing.